You are on page 1of 21

Simulation of Argon Gas Flow

Effects in a Continuous Slab Caster


B.G. THOMAS, X. HUANG, and R.C. SUSSMAN
Three-dimensional finite-volume-based numerical models of fluid, heat, and mass transport have
been developed and applied to help explain the complex inter-related phenomena of multiphase
fluid flow, superheat dissipation, and grade intermixing during the continuous casting of steel
slabs. Gas bubbles are simulated using a continuum model, which calculates the volume fraction
and velocities of the gas, and its effect on the liquid flow. Turbulence has been incorporated
using the standard K - e turbulence model. Reasonable agreement has been achieved between
predicted velocities and corresponding measurements and observations in full-scale water
models, both with and without gas injection. The effects of argon gas bubble injection on flowrelated phenomena are investigated with simulations of a typical steel slab caster. Argon bubbles
alter the flow pattern in the upper recirculation zone, shifting the impingement point and recirculation zones upward. The effect increases with increasing gas fraction and decreasing bubble size. Argon injection also causes superheat to be removed higher in the caster, moves the
hot spot upward, lowers the peak heat flux, and increases heat extraction from the wide face
and meniscus regions. During a steel grade transition, argon injection slightly affects slab surface composition but has no effect on intermixing in the slab interior.

I.

INTRODUCTION

A R G O N gas is employed at several stages in the continuous casting process (ladle, tundish, and mold) to encourage mixing, to help prevent nozzle clogging, and to
promote the flotation of solid inclusion particles from the
liquid steel. It usually enters the continuous casting mold
after injection into the submerged entry nozzle (SEN),
and eventually escapes from the liquid steel surface
through the mold flux powder layer.
Argon injected into the SEN provides a positive pressure inside the nozzle, which inhibits the natural aspiration of air through cracks, pores, or joints in the nozzle
walls, such as the junction between the nozzle and
sliding-gate flow-metering plates. This helps to prevent
the formation of the associated reoxidation products,
such as alumina, which otherwise form detrimental solid
inclusions that can adhere to the nozzle walls to block
the flow of liquid or enter the mold cavity with the steel
to create other quality problems. Argon helps to prevent
nozzle clogging in a second way by creating turbulence
that discourages adherence of inclusions to the nozzle
walls. Once inside the mold cavity, the argon bubbles
are believed to preferentially attach with inclusion particles, thus promoting their removal when the argon
leaves the surface of the liquid pool.
In addition to these beneficial effects, argon can have
detrimental side effects. Gas bubbles, and their associated clusters of attached inclusions, may become trapped
in the solidifying shell and create serious defects just beneath the slab surface. Alternatively, they may disturb
the liquid steel surface when they exit the powder layer,
B.G. THOMAS, Associate Professor, and X. HUANG,
Postdoctoral Research Associate, are with the Department of
Mechanical and Industrial Engineering, University of Illinois at
Urbana-Champaign, Urbana, IL 61801. R.C. SUSSMAN, General
Manager of Process Research, is with Armco, Inc., Middletown, OH
45043.
Manuscript submitted July 6, 1993.
METALLURGICAL AND MATERIALS TRANSACTIONS B

generating detrimental surface turbulence and consequent surface defects. These effects are influenced by
the amount and size of the argon bubbles and the flow
pattern inside the mold cavity. Moreover, the injected
argon gas bubbles also influence this flow pattern. In
addition, this has corresponding effects on the extraction
of superheat, composition intermixing during a grade
transition, and the movement of solid inclusion particles.
The extent of these effects is intensified by the volume
expansion of the gas bubbles in the high-temperature
molten steel. These effects have not been quantified previously, and they are the focus of the present modeling
study.
If.

PREVIOUS WORK

Experimental measurements on an operating continuous casting machine are very difficult, dangerous, and
expensive. Multiphase flow phenomena, particularly
volumetric expansion of the gas, are also difficult to simulate using physical water models. The recent development of numerical modeling provides an alternative tool
for understanding and solving this kind of problem in
material processing. Several mathematical models have
been applied to argon-steel flow and its associated heat
and mass transfer in gas-agitated vessels, such as casting
ladles, tl-m Research efforts have also been dedicated to
the fundamentals of gas-liquid flow dynamics and behavior of bubbles and bubble plumes, both with mathematical modelsf12-15~and with physical models using hot
liquid metals (liquid iron, steel, and copper) t~6ATl and
other liquids such as water, mercury, liquid silver,
butanol solute,
aqueous glycerol, and ethyl
alcohol.112-~5,17 24]

Relatively little work has been reported on two-phase


flow in the continuous casting mold. One of the few
studies, by Bessho et al., 125J compared the calculated
VOLUME 25B, AUGUST 1994--527

flow pattern, gas holdup (volume fraction), and inclusion distribution in a full-scale water model with experimental measurements and observations. Although
only one case was reported, the results showed that gas
created a great change in the flow pattern. A water modeling study observed the dependence of bubble penetration depth on nozzle design, c261 The authors concluded
that bubble dispersion can be controlled by nozzle submergence depth and nozzle geometry, such as nominal
port angle. However, the effect of gas bubbles on the
flow was not investigated directly. Another water modeling study, by Andrzejewski et al. ,t27j found that carefully controlled argon injection and submergence depth
were able to improve flow in a wide mold and even reduce surface-flow velocity and level fluctuations.
The present work describes the development of threedimensional (3-D) finite-volume models of two-phase
flow of liquid steel with argon gas bubbles and its associated heat and mass transfer in the continuous slab
casting mold. After verification with water models, the
models are applied to investigate the effects of bubble
size and injection rate on the flow pattern, superheat extraction, and intermixing during a steel grade transition.
In later work, the model will be extended to simulate the
movement of inclusion particles.
The transient nature of flow in the mold, which causes
problems such as surface turbulence, is known to be very
important to steel quality. However, as a first step toward understanding this complex behavior, the present
work assumes a steady-state flow pattern and investigates the influence of argon gas injection on this flow
pattern and related phenomena.

III.

GAS-LIQUID

Ov~
--=0

Oxi

OXjL

\ oxj

Ox# d

+ 4,

fgy = O,

5 2 8 - - V O L U M E 25B, AUGUST 1994

fgz = --O'gg

[4]

~z,ff = m + g,

[5]

K2

Iz, = q , p -

[6]

OK-O(l~-~x
pvj axj axi
E

Ovj Oxj

+ Cl

Oxi

o a k - c2

[3]

[81

where

G~ = ~ -~x,\-~xj + ,gxil
ci = 1.44,
trr = 1.0,

c2 = 1.92,

[91

c , = 0.09,

(r~ = 1.3

[10]

Bubble dispersion in the gas-liquid mixture due to turbulent transport and diffusion is modeled by the following transport equation for the continuum gas bubble
concentration, O'g:
[11]

where the diffusivity of gas bubbles in the mixture, Dg,


is assumed to be equal to the turbulent diffusivity of a
solute element with turbulent Schmidt number, Sc,, set
to 1; i.e.,
Dg -

/xt

pSq

/zt

[12]

No momentum equation is solved for the low-density gas


phase. Instead, the bubbles are assumed to have reached
their steady-state terminal velocity in the vertical direction relative to the liquid phase, vs,, by the time they
have entered the domain of the mold cavity. The three
components of gas velocity are thus simply

[2]

where i and j = { 1 , 2 , 3} = { x , y , z } and {xt, x2, x3} =


{x, y, z} in the Cartesian system; fg:, as described above,
is the extra force acting on the liquid due to buoyancy
of the gas bubbles:

fgx = O,

fz = g

The Reynolds number in the caster, based on the hydraulic diameter (Table I), always exceeds 10,000 even
far below the mold. This indicates that the flow is highly
turbulent everywhere. Thus, the standard K-e turbulence
model is used to calculate velocities of the liquid phase:

[1]

+ or ,,ov,+ ov,/1 +

fy = 0,

Oo'g _ 0 ( OtTg~
Vgj Ox~ Oxj Dg Oxj/

A finite-volume-based numerical model has been developed to simulate 3-D two-phase (gas-liquid) flows,
using the computational domain and grid of 60 x 34 x
16 nodes shown in Figure 1. Twofold symmetry is assumed, so only one quarter of the mold is modeled. The
buoyancy force acting on the liquid because of the gas
bubbles was taken into account by adding an extra force
term, fgz, into the momentum equation for the liquid
phase in the vertical direction, z. The liquid velocities,
vi (vi = {v~, vy, v~}), and pressure, p, are defined by the
3-D, incompressible, steady-state, mass and momentum
conservation equations for a Newtonian fluid:

Oxj

f~ = 0,

FLOW MODEL

A. Governing Equations

pv,

and f~ are body forces which include only gravity in this


work:

v~x = v~

[13]

Vgy = Vy

[14]

v~z = vz - Vg,

[15]

The gas bubbles are assumed to be spheres with a uniform size, whose terminal velocity is found according to
an empirical correlation tzS1developed for large individual
air bubbles rising in water:

Vg, = exp (a0) exp (al In dg) exp [a2 (In dg)2]
a0 = - 8 . 3 7 3 ,

al = - 2 . 6 3 0 6 ,

a2 = - 0 . 2 5 0 0

[16]
[17]

METALLURGICAL AND MATERIALS TRANSACTIONS B

Mold Wide.Face
(Cutaway)

.'US

rfa~
ween

dlfylng
,U and

Jld
;I

ver
In Caster
Fig. 1 - - S i m u l a t i o n domain and typical mesh used in 3-D steel-argon
two-phase flow model.

B. Boundary Conditions
1. Liquid phase
Inlet boundary conditions, including jet angle, velocity profile, turbulent kinetic energy and its dissipation
rate, are specified according to output from a separate
3-D finite-element model of the nozzle, described elsewhere, t29,3~ The inlet, shown in Figure 1, represents
the plane between the nozzle port and the mold cavity.
To simplify its geometry in the present mold model, a
rectangular inlet area is adopted for both rectangular and
circular nozzle ports.
Because fluid always flows out through the bottom
portion of the nozzle outlet port, and there is a small
inward flow near the top portion, [29,3~ the submergence depth used in the mold simulation refers to the
distance from the top surface of the mold to the top of
the jet, which is at the top of the inlet. This explains the
greater jet submergence depths, L,, employed to simulate the corresponding experiments, Lo. The distance between the top surface and the position of the maximum

METALLURGICAL AND MATERIALS TRANSACTIONS B

Outflow from the nozzle is chosen to be the same in the


simulation and the experiments.
For the outlet of the computational domain, which is
a horizontal plane across the steel caster, and holes in
the bottom of a physical water model, the normal gradients (O/On) of all variables, including vx, Vy, Vz, K, e,
and p are set to zero. The same boundary conditions are
used for each node on a symmetry centerplane, except
that the velocity component normal to the symmetry
plane is set to zero. The top surface is treated the same
as a symmetry plane, and the small variations in the liquid level due to motion of the free surface are neglected.
The bottom of the physical water model domain is
simplified to make it symmetrical. The bottom of the
standard water model, case A1 in Table I, was modeled
with four holes in the symmetric half of the domain instead of four holes on just one side, as used in the experiments. This simplification is reasonable because it
should have just a tiny effect on the flow only near the
outlet holes, and flow asymmetry was not observed in
the physical water model.
Empirical "wall law" functions t321are employed to define the tangential velocities, K and e, at the near-wall
grid nodes in order to account for the steep gradients that
exist near the walls. I331 When simulating steel casters,
the domain extends up to, but does not include, the
mushy zone or the solid shell. This avoids the computational difficulties associated with applying wall functions and modeling latent heat evolution at an internal
solid/liquid interface. The boundaries of the mesh along
the narrow and wide face walls correspond to the dendrite tips forming the outer limit of the mushy zone.
These boundaries are treated as very rough solid walls
by reducing the roughness factor in the wall laws from
8.8 to 0.8.1311 The validation of this approach of handling
the computational domain has been addressed in previous work. [30.33.34]

2. Gas phase
A zero-gradient condition of gas bubble volume fraction, ~rg, is set for all boundaries except the inlet of the
domain. This condition is consistent with no gas flow
through the walls, while it allows gas to leave from the
top surface, at the imposed relative terminal velocity.
Gas could also be carried from the bottom of the domain, if the terminal velocity did not greatly exceed the
casting speed.
C. Gas Bubble Size and Volume Fraction
at Mold Inlet
Gas bubble size and volume fraction are very important input parameters which control bubble behavior in
the flow simulation. Both of these parameters change
significantly after injection into the liquid steel in the
SEN as a result of both heating by the liquid steel and
pressure variation. In this work, a simple formula was
derived to calculate gas bubble size at the mold inlet
plane, dg, by applying the ideal gas state equation and
assuming bubbles are injected at standard conditions
(temperature 25 ~ and pressure 1 atm):

VOLUME 25B, AUGUST 1994--529

Table I.
Case AI
(51-mm 76-mm
Rectangular Port,

Armco)
L~ (ram)
Lh (ram)
a0 (~
a (~
L. (m)
L0 (m)
Z (m)
W (m)
N (m)
V~ (m/s)
E
P-0 (N s/m s)
p (kg/m 3)
v~ ( m / s )
v~ (m/s)
v~0 (m/s)
K0 (m2/s2)
eo (mZ/s3)
Gas
Q~ (m3/s)
O'~o(pct)
d~i (mm)
d~ (mm)

Simulation Conditions for Flow Model

Case A2*
(51-mm x 89-mm
Rectangular Port,
Armco)

Case A3*
(5 l-ram
Circular Port,
Armco)

46

36
25 down

51
66
15 down
25 down
0.1828
0.1524
2,152
1.93
0,229
0.0152
8.8
0.001
1000
1.048
0
0.489
0.0502
0.457
helium
0, 0.00022
0, 3 pet
1, 5
1, 5

0.1778

1.513
0
0.672
0.0281
0.705
none
0
0
none
none

1.908
0
0.848
0.0702
1.335
none
0
0
none
none

Case B1
(Long Model Inland)

Case B2**
(Short Model Inland)

60
38
15 down
25 down
0.265
0.235
3
1.32
0.22
0.0167
8.8
0.001
I000
1.062
0
0.427
0.054
0.447
none
0
0
none
none

1.12

air
0.00026
5 pet
1
1

Case C
(Steel Caster)
60
38
15 down
24 down
0.265
0.235
3
1.32
0.22
0.01667
0.8
0.0055
7020
t .062
0
0.471
0.0502
0.457
argon
0, 0.00012, 0.00026
0, I 1, 22 pct
0.6, 1.7, 3
1,3,5

*Unlisted values are the same as case A1.


**Unlisted values are the same as case B 1.

118]

kPg/
P~-2/= (

P-~pgl~)(T~

[19]

where &~ and & are the gas densities at the injection
point and the nozzle port, respectively; P= is the ambient
pressure (1 atm). The bubble temperature at the mold
inlet is set to that of the liquid steel in derivation of
Eq. [19] because the bubbles heat very quickly, as
shown in the Appendix. It should also be noted that pressures at the flow meter and at the injection point in the
SEN are likely to be higher.
Next, applying a mass balance on the gas flow
through the system, the gas volume fraction at the mold
inlet plane, 0-80, can be obtained:

Qg pg._.~i
~g0 :

Pg
Qe pe--2i+ VcNW( t - ~r~o~t)
Pe

[201

Equations [18] through [20] are used to determine the


bubble size and gas volume fraction at the mold inlet of
the steel caster, given the initial bubble size, dgi, and
volume flow rate, Qg, and assuming the outlet volume
fraction, ~gou,, is negligible. For water model simulations, the density remains constant, so the gas bubble
size remains constant, and the volume fraction from
Eq. [20] simplifies.
530--VOLUME 25B, AUGUST 1994

D. Solution Method
Owing to the simple rectangular geometry of the
mold, a computer code based on finite-difference calculations, MUPFAHT I35J has been chosen for this complex problem. The steady-state (elliptic) system of
differential equations and boundary conditions is discretized into finite-difference equations using a staggered
grid and seven-point stencil of control volumes. To aid
convergence, an upwinding scheme is employed for the
advection terms in domains with high cell Reynolds
number, t361 In addition, the source terms are linearized
to increase diagonal dominance of the coefficient matrix. t36~ The equations are solved with the semi-implicit
method of pressure-linked equations algorithm, whose
alternating-direction-semi-implicit iteration scheme consists of three successive tri-diagonal-matrix-algorithm
solutions (one for each coordinate direction) followed by
a pressure-velocity modification to satisfy the mass conservation equation. [36}
Obtaining reasonably converged velocity and turbulence fields for this problem is difficult because of the
high degree of recirculation. The current strategy employed is successive iteration using an under-relaxation
factor of 0.2 or 0.3 until the maximum relative residual
error and maximum relative error between successive solutions fall below 0.1 pct. Based on this criterion and
starting from an initial guess of zero velocity, over 2500
iterations are required to achieve a converged solution.
This takes about 20 CPU hours on a Silicon Graphics
4D/35 workstation for a mesh of 60 34 16.
IV. H E A T - T R A N S F E R M O D E L
FOR S U P E R H E A T D I S S I P A T I O N

The dissipation of the superheat has an important influence on growth of the shell during the critical initial
METALLURGICAL AND

MATERIALS

TRANSACTIONS 1B

stages of solidification. It also has an important effect


on surface defect formation and steel internal quality
related to the microstructure. To investigate superheat
dissipation in the continuous slab casting mold, a heattransfer model has been developed to compute temperature distribution within the liquid pool, heat transfer to
the inside of the solidifying shell, and its effect on
growth of the shell. This model solves a 3-D energy conservation equation:

pep Yj

: OXj ~k eff OXj/

[21]

Because conductive heat transfer is enhanced greatly


by turbulent eddy motion, the effective thermal conductivity, k~ff, consists of two components:

Cp~,

k~ff = k0 + - Prt

[22]

k~efdepends greatly on the turbulence parameters through


the calculated turbulent viscosity, /x,, and the turbulent
Prandtl number, Pr,, which is set to the standard value
of 0.9 in the present work. ~3~

thermal stress and shrinkage model that includes coupled


heat flow across the mold/shell gap. 139]

B. Incorporation of Gas Bubbles


Argon gas bubbles influence heat transfer in the caster
in two ways: by altering thermal convection through
their effect on flow velocities, and by enhancing turbulent heat conduction by increasing turbulence intensity. This effect of gas on heat transfer has automatically
been taken into account through the two-phase flow
model described in Section III.

C. Solution Methodology
The differential equations, together with the boundary
conditions described in Section A, were solved with the
same finite-difference schemes used for the flow
models. Because heat transfer has negligible effect on
the fluid flow, (i.e., there is only one-way coupling between the fluid-flow and heat-transfer models), previously converged solutions of the velocity and turbulence
fields are input and kept unchanged during the solution
of the temperature. Thus, only 30 minutes CPU time is
needed on the Silicon Graphics 4D/35 workstation.

A. Boundary Conditions
Temperature across the inlet plane is simply fixed to
the casting temperature, To. This temperature corresponds to a tundish temperature, because the temperature drop through the nozzle is very small. I371 Adiabatic
conditions or zero normal temperature gradient conditions (OT/On) are used at the outlet plane and the symmetry centerplanes. For the top surface, calculations
were made to estimate heat conduction through the molten flux and powder layers and radiating to ambient, t3~
To account for this heat loss, an equivalent thermal convection boundary condition is applied to the top surface,
using the heat-transfer coefficient, h, and ambient temperature, T~.
To correspond with the flow boundary conditions, the
boundaries of the computational domain for heat transfer
are again assumed to be the dendrite tips, so the domain
does not include the solidified shell and the mushy zone.
A fixed temperature, nominally equal to the liquidus,
Tliq, is imposed along these boundaries. The reduction
of the domain due to the solidification is neglected because the solidifying shell is very thin in the mold
region.
An empirical "thermal wall law "f33~ is used to determine temperature at the near-wall grid nodes. Use of this
thermal wall function was important to achieve an accurate heat balance. It is needed to calculate the heat flux
due to superheat dissipation, qs~, which in turn influences the growth of the solidifying steel shell.
This approach differs from other recent models, which
couple the fluid-flow and solidification calculations. The
latter models use a function (based on flow through porous media) to radically reduce velocity and turbulence
levels within the mushy zone.t38~ By separating the fluidflow and solidification calculations, the present approach
reduces the complexity needed in subsequent models of
heat conduction and solidification of the shell. Results
from the present model have been incorporated into a
METALLURGICAL AND MATERIALS TRANSACTIONS B

V.

MASS-TRANSFER MODEL
FOR GRADE TRANSITION

A 3-D transient mass-transfer model, consisting of


three submodels, has been developed to calculate intermixing in both the strand and the solidified slab during
a transition in steel grade, t4~ The results presented in this
work correspond to a "flying tundish change," where the
steel grade is changed simultaneously with the ladle and
tundish. The influence of argon gas bubble injection on
mass transfer is automatically introduced through its effect on the time-averaged velocities, as done with heat
transfer.

A. 3-D Transient Mass-Transfer Model


of Upper Strand
The first submodel calculates 3-D transient turbulent
mass transfer of solute in the upper 6 m of the strand by
solving the 3-D transient transport equation
--

Ot

+ vj-

Oxj

Oxj

Dcfe

[23]

where C is the dimensionless composition, or "relative


concentration," defined by
C =-

F(x, y, z, t) - Fold

[24]

F,~,~ - Fold
and F(x, y, z, t) is the fraction of a given element at a
specified position in the strand or slab; Fold and F.r are
the desired fractions of that element in old and new
grades respectively; Deaf is the effective diffusivity given
by
/z,
D~ff = Do + ~

pSc,

[25]

VOLUME 25B, AUGUST 1 9 9 4 - - 5 3 1

The initial value of C is set to zero throughout the


domain to start the simulation. A sudden transition from
C = 0 to C = I is imposed at the inlet plane, (which
defines z = 0 in the final slab). Zero-gradient or "no
mass diffusion" boundary conditions were applied to the
composition at all other boundaries. Thus, solute can
only leave the domain by fluid transport across the outlet
plane.

properties with the flow model. The solution method


employs a backward Eulerian iteration within each
variable-size time-step. A simulation of 960 seconds of
casting requires about 50 time-steps and 8 hours of CPU
time on the Silicon Graphics 4D/35 workstation. Further
details are provided elsewhere, t4~
VI.

B. 1-D Mass Transfer Model of Lower Strand


Composition evolution must be calculated within the
entire liquid pool of the strand (usually 20 to 40 m to
the metallurgical length) before it is possible to predict
the complete composition distribution in the final slabs.
Fortunately, the 3-D results for the top 6 m indicate that
the velocity profile in the lower region of the strand
(below 6 m) is quite uniform and close to that of turbulent flow through a duct. Thus, for economy, a 1-D
mass-transfer model of Eq. [23] in the z direction was
developed as the second submodel, to simulate the remaining domain beyond 6 m.

TYPICAL MODEL FLOW RESULTS

The 3-D numerical model of two-phase flow, described in Section III, can simulate flow in either the
actual steel slab continuous casting machine, or in a
physical water model of the process, by making simple
changes in the domain outlet boundary conditions, the
liquid properties, and other conditions as desired. This
section presents typical predictions of the flow pattern
in a 2.15-m-high water model, for conditions A1 in
Table I. The results are later compared with observations
and measurements in the physical water model. The
post-processor FIPOST, of the commercial finiteelement program FIDAP, ~41] was used to visualize and
plot the results.

C. Slab Composition Model


Composition distribution in the final slab develops as
the solidifying shell grows in thickness down the caster.
The third submodel calculates the composition distribution in the final slab from the 3-D time-varying concentration history of the strand, generated by the first
two submodels. Composition at each point in the strand
is assumed to evolve according to the calculated history
until that point solidifies. Diffusion in the solid and
macrosegregation are ignored. Mathematically, this
model performs a coordinate transformation on the
strand results, C, to obtain the composition in the final
s l a b , Cslab:

Cslab(Xs, y,, z,) -: C(x, y, z, t)

[26]

where the spatial and time coordinates in the strand, x,


y, z, and t, are related to the coordinates in the final slab,
x,, Ys, z~, through the following equations:

x = x,,

y = y,,

z = fots v,dt,

t = f zdz
--

[27]

s lflz

In the preceding equations, t, represents the solidification time of the steel shell, which defines the depth
(z distance) beneath the strand surface where the composition no longer changes. To specify t,, a simple relationship involving x-y position in the strand, shell
thickness, AL, and solidification constant, k,hez~, was
adopted:
t, =

AL = rnin

- Y'

Figure 2 illustrates the typical 3-D flow pattern predicted numerically for a water model with no gas injection. The actual water model will be described in the
Section VII. For clarity of presentation, a velocity vector
is drawn only at every eighth node in the grid.
Figure 2(a) views the centerplane section parallel to the
wide face wall. The fluid leaves the nozzle as a strong
jet, impinges upon the narrow face, then splits vertically
to create upper and lower recirculation regions.
Figure 2(b) reveals the interior velocity vectors in transverse sections taken at five locations down the mold.
Because the bifurcated nozzle sends flow into a relatively thin mold cavity, the resulting velocities are relatively uniform through the thickness of the mold over
most of the mold interior. Down the corner near the impingement point, a weak vortex is formed, as the jet
spreads across the narrow face and meets the incoming
flow just off the corner along the wide face surface. Velocity components through the mold thickness are quite
small everywhere except very near the outlet holes.
Thus, flow in the mold can be characterized by the angle
of the jet traversing the mold, the location of its impingement point on the narrow face, and by the location
in the x-z plane of the centers of the recirculation zones:
the "lower eye" and "upper eye. ~ These flow parameters
are illustrated in Figure 3.

B i Typical Flow Predictions in Water Model


with Gas Injection

x
[28]

D. Solution Method
Like the solution procedure for the heat-transfer
model, the 3-D and 1-D transient transport equations are
solved after first obtaining the velocities and turbulence
532--VOLUME 25B, AUGUST 1994

A. Typical Flow Predictions in Water Model


without Gas Injection

The influence of gas bubbles on the flow pattern is


seen in Figures 3(b) and 4 for a uniform gas bubble size,
dg, of 1-mm diameter and a gas volume fraction at the
inlet plane to the mold cavity, ~rgo, of 3 pct. The buoyancy of the gas bubbles changes the flow pattern significantly in the upper region of the mold, even for this
small amount of gas injection. When the bubbles are
METALLURGICAL AND MATERIALS TRANSACTIONS B

Model
Cenmd~n~

"s

0.5 m/s

Model
Centerline
0.5 m/s

0.20

..o
8

0.40
0.35-

0.53-

(a) 3-D view of velocity


vectors.

(b) Velocity vectors in transverse


sections down water model.

Fig. 2 - - P r e d i c t e d flow patterns in water model (case AI in Table I without gas).

0.5 m/s
0

~ 0.5 m/s
Upper Eye
J
................

........

\
o
,~

,,,,

......

,~l~

" ........

IIItlllll,

' ........

.......

Illll|

tlll||~

N,''"
I t l l l a l ~ , , , , . . . . ~ t Jllllll]
..................
,,l~|l ]

l. . . . .

i;iZ:~'. ''''''j"tt

J. . . . . . . . .
I~I~
I. . . . . . . .

~"

7"' ......

1.5 . . . . . . . . . . . . . . . . . . . . .
IIII

%~1%%~%

. . . .

"lllllllll~J

,..~.

,. i iiiiiiiiiiiiiiiii

.....................
, ,,,,,,,..
., ... .. ... .. ... .. ... ..... . . . . . . . . ..~...l l / l l l n , ~
.................
. . . ,, ,, ,, , , . , .

zo- iiii

'tl|lllt

,I

. . . . . . . . ",l~t~
I l llll,
I'lll,I ~,,IISSS . . . . . . . . .
IIIILI
..............

:7.

::::::::::::::::::::::::::::

i i ! ! ,Ii
( a ) No gas.

( b ) With gas (bubble size


1 ram, 3% gas at inlet).

Fig. 3 - - T w o - d i m e n s i o n a l view of predicted flow patterns in water


model (case A1 in Table I).
METALLURGICAL AND MATERIALS TRANSACTIONS B

small, as assumed here, they travel with the jet and induce it to bend upward to impinge at a slightly higher
location on the narrow face wall. This upward motion
due to the bubbles is predicted to shift the location of
the upper eye greatly toward the mold center and slightly
upward to just above the nozzle port.
As the bubble-rich liquid in the central portion of the
jet rises vertically, velocity gradients are created through
the mold thickness. This effect is seen most clearly by
comparing the 0.20-m slices in Figures 2(b) and 4. A
large horizontal recirculation region is evident, so there
is no unique location of the upper eye (Figure 4). As the
rising bubble-rich liquid reaches the top surface, it slows
down the flow of liquid across the top surface back toward the nozzle by 50 pct and creates a slight drift toward the wide faces. Both of these effects disrupt the
mainly 2-D flow pattern found without gas. These predictions agree with the observations in water models reported by others. {27]
Gas bubbles are predicted to have much less effect on
the flow pattern in the lower portion of the mold. The
lower recirculation zone appears very similar to that
without gas, with the lower eye raised only a small distance. The reason for this diminished effect is explained
in Figure 5, which shows the predicted gas volume fraction for these conditions. Because of the strong buoyancy force, which is proportional to the density
VOLUME 25B, AUGUST 1994--533

Model
Centerline

"s

Model
Centerline

0.20

-~.

Contour Unit : %

0.5 m/s

0.5
0.40

.8
0.35

1.0

0.53.

2.0

Fig. 4 - - V e l o c i t y vectors in transverse sections down water model


(case A1 in Table I, bubble size 1 mm, 3 pct gas at inlet).
Fig. 5 - - P r e d i c t e d gas volume fraction in water model (case A1 in
Table I, bubble size 1 mm, 3 pct gas at inlet).

difference between liquid and gas, most of the gas bubbles float upward and escape from the top surface during
the time that the liquid jet takes to travel from the nozzle
port to the narrow face wall. Less than 5 pct of the bubbles ever reach the vicinity of the narrow wall, and even
fewer enter the lower recirculation zone of the liquid
flow. Thus, flow in the lower recirculation zone should
be relatively unaffected by gas bubble injection.

C. Effect of Gas Bubble Model


The influence of the gas bubbles on the liquid has
been incorporated into the model solely through the
buoyancy force created by the rising bubbles. There are
at least two other factors, ignored in the present model,
which are worthy of further discussion. The first is the
effect of the momentum imparted from the bubbles to
the liquid. This is expected to be negligible because of
the small bubble size combined with the very small gas
density relative to the liquid. The second effect is the
reduction in the volume fraction of the liquid due to the
presence of the gas bubbles. A preliminary study incorporating this effect into the model revealed that correctly
decreasing the liquid volume increases liquid velocities
near the inlet port. Elsewhere, however, there are no significant changes in either the velocities, temperature predictions, or agreement with experiments,

534--VOLUME 25B, AUGUST 1994

VII. P H Y S I C A L W A T E R
MODEL EXPERIMENTS

To verify the numerical flow model predictions, measurements of velocity profiles and flow pattern observations have been made with full-scale water models at
Armco Research Center (Middletown, OH) and Inland
Steel (East Chicago, IN). The 1.93-m- (76-in.-) wide
and 0.229-m- (9-in.-) thick "water caster," shown schematically in Figure 6, is a clear plastic representation of
an actual slab caster used in the Armco Middletown
Works, with its length shortened to 2.15 m. This physical model has four pipes located at the bottom of the
wide face to allow removal of water at a volume flow
rate corresponding to the casting speed. Helium gas was
injected into the slide gate on some trials to simulate
argon injection through the submerged entry nozzle at
the steel caster.
The flow patterns are visualized in three ways:
(1) blue ink injected as a pulse from the tundish slide
gate; (2) observation of a flag constructed with thin film
attached to a wooden dowel; (3) helium gas bubbles injected through the SEN. Observations using these three
methods are combined to estimate the average downward angle of the jet traversing the mold, the impingement point of the jet against the narrow face, and the

METALLURGICAL AND MATERIALS TRANSACTIONS B

Water with Gas

Liquid

~unace

""

\*
_ \

J,

f " " - ~ T o Voltage Supply and

1.2

sign=proces.~r/

~" 0.9
"

0.6

(I 0

10

12

14

16

18

Tlme (s)
Fig. 7 - - T y p i c a l signals output f r o m the strip chart record (mean
velocity = 0.23 m / s , standard deviation = 0.08 m/s).
B- B : ProbeTIp

to "wander" around with time, producing a lowfrequency variation in the velocity signal of 2 to 5 seconds. To account for this effect and to ensure that a true
time-mean speed measurement was obtained, over
2 minutes of signal measurements were averaged to obtain each speed data point.
Wall

Outlet PI

VIII.
Valve

To Pump

Fig. 6 - - S c h e m a t i c of physical water model and hot-wire anemometry apparatus for speed measurements.

locations of the centers of the upper and lower recirculation zones.


A hot-wire anemometer probe with a single-wire sensor is used to measure the velocity profiles. Carefully
measured resistance heating power is provided to the
wire to balance the heat loss, which increases in proportion to the flow velocity component perpendicular to
the wire. By orienting the sensor wire perpendicular to
the wide face wall, this probe measures the speed component in the wide face plane. The calibrated voltage
signals from the sensor are recorded by both a strip chart
and a needle, whose movements are dampened electronically to produce a partially time-averaged signal.
After careful calibration of the system, including an
adjustment for water temperature, the probe was traversed manually across the mold at the symmetry plane
through the narrow face of the water model. Measurements were taken at 50-ram intervals down each vertical
line from the top surface. Extra data points were measured near the nozzle port and impingement point, as
determined by flow pattern visualization, because large
velocity gradients were expected in these regions.
A typical signal output from the strip chart recorder is
shown in Figure 7. For this example, the time-averaged
speed is 0.23 m / s , and the standard deviation, containing 68 pct of the signals, is about 0.08 m/s. Note that
the maximum range of the signal is -+0.18 m / s , indicating the tremendous effect of the time-dependent turbulent fluctuations of the flow field. The jet is observed

METALLURGICAL AND MATERIALS TRANSACTIONS B

NUMERICAL MODEL VALIDATION

To validate the accuracy of the numerical flow model,


described in Section III, simulations were run to match
the flow conditions in the full-scale physical water
models described in Section VII. This section compares
the numerical predictions with the experimental observations of flow pattern, gas bubble distribution, and
speed measurements.

A. Comparison with Flow Pattern Observations


The overall flow pattern calculated numerically is very
similar to the flow pattern observed in the physical
models. Table II quantitatively compares the flow parameters for simulation conditions, case AI in Table I,
matching experiments performed at Armco Research.t421
For the comparisons with gas, the gas volume fraction,
o-g0, of 3 pct (corresponding to 0.00022 m3/s), was chosen to approximate the flow rate of 0.5 SCFM (standard
cubic feet per minute) of helium injected into the physical model nozzle through a small tube of inner and outer
diameter of 4 and 6.4 mm, respectively. The bubble size
of 5 mm was estimated both by visual observations and
by examining photographs of the bubbles in the water
model.
The five measurements of eye and impingement point
location agree with the numerical predictions both with
and without gas. The largest discrepancy is the slightly
deeper lower eye predicted by the numerical model. This
might be because the jet in the physical model achieves
fully developed turbulent channel flow more rapidly than
in the computational model cavity. Consequently, the recirculation zone in the physical model is shortened.
The flow predictions agree with the measurements that
the effect of the gas bubbles on the flow pattern is quite
small, for this case with only 3 pct gas injection. As seen
in Table II, gas bubbles consistently raise the depth of

VOLUME 25B, AUGUST 1994--535

Table II.

Experimental and Predicted Eyes and Impingement Point in Water Model (Case AI in Table I)

No Gas
Predicted
Experimental
Lower eye:
Distance to centerline (m)
Depth (m)
Upper eye:
Distance to centerline (m)
Depth (m)
Impingement point depth (m)
Jet angle (~
Surface speed (m/s)
(midway between
nozzle and narrow face
at centerplane)

Experimental

0.521
1.143

0.55
0.80

0.521
1.041

0.53
0.76

0.521
0.254
0.498
25 down

0.55
0.20
0.50
28 down

0.579
0.198
0.442
25 down

0.53
0.24
0.48
29 down

0.23

0.25

the impingement point and the lower eye. Although


small, the direction of movement of the upper eye does
not agree. This might be due to the difficulty in measuring this location, which wanders with time and is not
uniform through the thickness of the water model.

B. Comparison with Speed Measurements


Predicted velocity profiles are compared in Figure 8
with experimental speed measurements on the Armco
water model for simulation conditions corresponding to
cases A2 and A3 in Table I. This figure shows an agreement between predicted and measured velocities, both
qualitatively and quantitatively. The results show how
the jet spreads as it moves across the mold. Its peak velocity decays to about 30 pct of the inlet value by the
time the jet is halfway to the narrow face wall. Near the
narrow face wall, the velocity profile has a concave
shape, due to stagnation at the point of impingement.
Note in Figure 8(b) that speeds measured earlier (solid
circles) I43j are larger than recent measurements (open circles), which agree more closely with the numerical
predictions.
No major differences are seen between the velocities
produced by the different nozzle port shapes compared
in Figures 8(a) and (b). The characteristics of the rectangular and circular nozzle ports compared here are
given in cases A2 and A3 in Table I. The separate 3-D
model of nozzle flow found that these two nozzles produce jets with a similar angle, outlet area, and flow rate,
despite differences in their nominal geometries, t291 Both
the mathematical and physical models show that the two
ports produce similar velocity profiles in the mold as
well, except for a small difference in magnitude, which
is reproduced by the calculations. This implies that the
steady flow pattern in the mold is controlled solely by
the angle, outlet area, and flow rate of the jet exiting the
nozzle. Thus, nozzle port shape alone should have no
significant effect on the flow pattern and related phenomena. This might appear to contradict other work,
which has reported an important influence of nozzle port
shape on turbulence at the meniscus. 144J However, the
present work concludes only that the nozzle port shape
5 3 6 - - V O L U M E 25B, AUGUST 1994

With Gas
Predicted
(Size 5 mm, 3 pct Gas)

0.11

alone has no effect on the time-mean flow pattern. Nozzle shape might affect the casting process through its influence on nonsteady phenomena, such as time
variations in turbulence or asymmetric surging, which
require further study.
Figure 9 shows the similarity between the predicted
and measured effects of gas bubble injection on the
speed profiles down the mold interior. As observed visually, a 3 pct injection of 1-mm-diameter helium bubbles has a relatively small effect on the jet. Predictions
and measurements both show that gas injection widens
the jet slightly and diminishes its peak. Quantitatively,
the measurements are consistently higher than the numerical predictions, as expected from the discussion of
Figure 8(b).
Reasonable agreement between predicted and measured velocities is seen again in Figure 10. The experimental data shown in this figure were obtained at Inland
Steel 1451 and correspond to the simulation conditions
given in case B 1 of Table I. The velocity stagnation at
the impingement point, which traverses speed only
5 mm from the narrow face wall, is particularly evident
in this figure, both experimentally and numerically. The
stagnation point falls between adjacent velocity peaks,
where the jet splits to flow upward and downward along
the narrow face. These measurements also show that the
speed on the right side of the water model is higher than
the left side, due to the asymmetrical opening of the nozzle slide gate. t45~ This flow asymmetry was not accounted for in the numerical model results reported here.
Nevertheless, the numerical model is able to reproduce
the main flow characteristics, including the locations of
impingement point and velocity peaks. Speed predictions match the measurements down the left wall quite
closely. Note that relatively little difference is predicted
between speeds at 7 and 14 mm from the wall.
It is interesting to note a slight overprediction of the
measured velocity between 1 and 2 m below the meniscus. This has been postulated to arise from the numerical
model's neglect of strand curvature, which was present
in the physical model. As the strand curves away from
the vertical nozzle, the greatest intensity of the jet down
the narrow face wall should move closer to the outer
radius. This should leave lower speeds at the center
METALLURGICAL AND MATERIALS TRANSACTIONS B

.........

, ....

,.

O~0

' ' ' 1 ' ' ' ' 1 ' ' ' ' 1 ' ' ' ' 1 "

',

"1

....

I ' ' ' ' 1

....

I"

:s

0.4
51 mm from
Nozzle Port
0

0.6
....

I~

460 mm from
Nozzle Port

921 mm from
Nozzle Port

Pred.
Exp.

o
L,,,I

102 mm from
Nozzle Port

I ....

| , . t , I .

9 ,,I

....

I , , , , I

....

i,

2o

,I

20

Speed (m/s)

....

I ....

I ....

Speed (m/s)

I,

I , , I , I

....

20

I ....

I ....

Speed (m/s)

I,

Speed (m/s)

(a)
0.0

9
" '1.'

li!,'
.... ' ........
o
.

'

o.2
Io

0.4

_m_

51 mm from

l0

--epo

,,,I

I e e / ~ 460 mm from

Pred,
Exp.
Exp., Armco

o
9

0.6

102 mm from

....

I ....

I ....

I . . . .

921 mm from

9
,*
I ....

. . . . . ... . . .

20

Speed (m/s)

20

Speed (m/s)

iL,k:.,

I ....

I ....

Speed (m/s)

I,

20

1 ....

I ....

I , , , , 1 ,

Speed (m/s)

(b)
Fig. 8--Comparison of calculated and measured velocity profiles in water model (cases A2 and A3 in Table I). (a) 51-mm (l-in.) by 89-mm
(3.5-in.) rectangular nozzle port (case A2 in Table I) and (b) circular nozzle port with a diameter 51-mm (2-in.) (case A3 in Table I).

plane, where speeds are compared in Figure 10. Strand


curvature was expected to cause a difference between
predicted and measured speeds, but the actual difference
was surprisingly small, which indicates that the flow
tends to bend to follow the strand curvature.
The greatest discrepancies between speed measurements and calculations exist in the low-velocity areas of
the mold. These include the velocity valley corresponding to the upper eye in the 460-mm frames of Figure 8,
and the valley near the impingement point in the 10-mm
frames of Figures 8 and 10. Measured values in these
regions are somewhat higher than the predictions for
several reasons. First, it is very difficult to measure the
flow near separation, reattachment, or impingement
points, which have low time-averaged velocity but high
turbulence levels. The eyes and impingement points
were observed to move around with time, so their locations are difficult to specify exactly. Second, the
single-sensor probe used in these experiments measures
only speed (velocity magnitude) and cannot detect reversals of flow direction, such as caused by turbulence.
M E T A L L U R G I C A L AND MATERIALS T R A N S A C T I O N S B

Because speeds are always recorded as positive, larger


time-averaged speeds are measured in these regions
where rapid flow reversals are common. The same observation has been found for flow in nozzles. [29] Finally,
there may be large-scale transient effects that cannot be
detected with the steady-state calculations.

C. Comparison with Gas Bubble


Distribution Observations
Accurate calculation of bubble distribution is important because gas bubbles affect the liquid flow pattern in
proportion to the calculated gas volume fraction, O-g, in
the present model. The calculated gas volume fractions
and corresponding flow pattern are compared in
Figure 11 with those observed at Inland Steel [3~ under
conditions B2 of Table I. The model predicts the same
tendency for gas bubble movement as shown in the
photograph. Most of the bubbles are crowded together
in the upper region of the mold cavity. After entering
V O L U M E 25B, A U G U S T 1 9 9 4 - - 5 3 7

O.Oi

. . . .

. . . .

i - - ' - . |

. . . .

the mold cavity with the liquid jet, the bubbles float
quickly upward through the recirculation zone, and leave
the top surface at their assumed terminal velocity, vg,,
(0.12 m / s for 1-mm bubbles). Fewer and fewer bubbles
stay in the jet as it travels across the mold. This figure
also shows that the flow pattern predictions are reasonable, and it illustrates how the bubbles buoy the liquid
jet slightly upward.
The model somewhat underpredicts bubble penetration to the deeper regions of the mold cavity. Several
reasons are suspected: (1) stronger turbulence in the
physical water model gives rise to higher turbulent dispersion than that predicted; (2) real, nonspherical bubbles have higher drag, so they are carried further by the
liquid jet; (3) bubbles do not leave the top surface at their
terminal velocity, as assumed by the model. Instead,
they collect at the meniscus under the action of surface
tension forces, which likely prolongs their average residence time.

. . . .

:~ 0.2
m

Nozzle Port
Pred.
9 Exp., Armco

~0.4
r

9 . . i

0.0

. . . .

. . . .

. . . .

102 mm from
Nozzle Port

}/"
t k . . , I

i , . .

1.0
2.0
Speed (m/s)

. . . .

0.0

. . . .

I . . . .

1.0

1 . . .

2.0

Speed (m/s)

(a) Without gas.


o oH

.............

. .......

F\
"~

[ /
[ /

04 !
I

Pred.

i . . . i

2 m m from
Nozzle Port

Nozzle Port

II

D. Verification of Heat- and Mass-Transfer Models

51 mmfrom
Exp.,Armco

. . . .

0.0

The heat- and mass-transfer models developed and described in this work have been compared favorably with
available data from plant trials, including measured liquid temperature, [37] shell thickness, [46] and slab composition, t42,47] Further discussion of the details of these
comparisons, as well as model results and parametric
studies, is given elsewhere. [33,4~

. . . .

. . . .

1.0

ii

r . . , i , . . . i

o.o

2.0

Speed (m/s)

(b) With

....

i . . i f l . . .

1.0
2.0
Speed (m/s)

gas (bubble size 1 mm, 3% gas at inlet).

Fig. 9 - - C o m p a r i s o n of calculated and measured velocity profiles


near nozzle port in water model (case A2 in Table I),

, NI ....

IT,,,I

....

oB
0

0.5

%0 o

,
v

E /

....

0 ~

!
O i l-"

2.5

0.05

[]
O

0.1

0.15

Pred., 14ram from NW


Pred., 7 mm from NW
Exp., Left
Exp., R~ght

0.2

0.4

0.25

0.3

0.35

Speed (m/s)
Fig. 1 0 - - C o m p a r i s o n of predicted and measured velocity profiles
near narrow wall in water model (case B 1 in Table I).
538--VOLUME 25B, AUGUST 1994

The extensive validation with water models, presented


in Section VIII, generates some confidence that the numerical models are capable of making reasonable predictions of multiphase flow, heat transfer, and mass
transfer, in the mold region of actual continuous steel
slab casters. The models are next applied to investigate
the effects of argon gas injection on these phenomena
under the typical casting conditions listed in Tables I
(case C) and III.

A. Bubble Expansion

1.5
o

IX. E F F E C T O F A R G O N GAS
I N J E C T I O N IN S T E E L C A S T E R

The gas volume fraction and bubble size at the inlet


(corresponding to the nozzle port) were determined by
considering the significant volume expansion that should
occur as the bubbles heat up while traveling down the
nozzle. This effect magnifies the importance of argon
gas in a steel caster, compared to that in a water model.
The bubble expansion and the corresponding increase of
gas volume flow rate into the mold are calculated according to the equations in Section II and data in case
C of Table I. The temperature rise almost doubles the
bubble diameter, which increases the gas volume flow
rate at the nozzle outlet to about five times higher than
that injected into the nozzle at standard conditions (25 ~
and 1 atm). Setting the gas volume flow rate into the
nozzle inlet at standard conditions to be 0.00026 m3/s
(0.55 SCFM or 5 pct) and expanding 3-mm bubbles by
a factor of 5.2 produce a 22 pct volume fraction of 5-ram
bubbles at the inlet to the mold (case C of Table I). The
METALLURGICAL AND MATERIALS TRANSACTIONS B

(a) Gas volume fraction


(contour unit : %)

(b) Liquid velocities


(---) 0.3 m/s)

Fig. 11--Comparison of calculated and measured velocities and gas volume fraction in water model (case B2 in Table I).

Table III. Simulation Conditions


for Heat-Transfer Model**

Cp
h
k0
Pro
Prt
Tli q

To
Tsol
T=
ATs

680 J kg J K-I
40 W m -2 K -1
26 W m -1 K -1
0. I
0.9
1525 ~
1550 ~
1518 ~
27 ~
25 ~

* * U n l i s t e d v a l u e s are the s a m e as c a s e C o f T a b l e I.

volume increase due to heating is over 10 times more


important than the volume change due to t h e pressure
change. Moreover, calculations in the Appendix show
that this heating is extremely rapid and is completed well
before the bubbles enter the mold cavity. Thus, it is reasonable to ignore the effects of temperature and pressure
changes on bubble behavior in the mold cavity itself.
The argon flow rates of 0 to 22 pct with sizes of 1 to
5 m m at the mold inlet were chosen in this study to represent the typical range of gas injection practices encountered in different casting operations.

B. Flow Pattern
Argon gas injection affects the casting process in part
through its influence on the liquid flow pattern. The extent of this effect depends on both the gas injection rate
METALLURGICAL AND MATERIALS TRANSACTIONS B

and the bubble size. Figures 12 and 13 show the effects


of these two important parameters on the liquid flow patterns and the movement of gas bubbles themselves, for
typical casting conditions. In contrast to flow in the
water model, 0.00026 m3/s gas injection into a steel
caster produces a substantial change in the flow pattern,
as the gas fraction entering the mold increases from 5 to
22 pct as a result of the gas expansion discussed in
Section A. When argon is present, the upper recirculation zone shrinks to a very small region near the nozzle
and may even disappear altogether. Flow across the surface of the mold completely reverses, as increased flow
toward the surface takes liquid back toward the narrow
face. The lower recirculation eye and impingement point
shift upward markedly. Figure 13 shows that most bubbles are concentrated in the central portion of the wide
face. More than 25 pct of the bubbles are predicted to
impinge upon the upper portion of the wide face walls,
during their upward flotation.

1. Effect of gas injection rate


Figure 12 shows that increasing the gas volume flow
rate entering the mold produces a corresponding increase
in its effect. Stronger buoyancy due to the introduction
of more bubbles not only changes the flow pattern to a
larger extent, but also makes the bubbles themselves
float more easily, resulting in shallower bubble dispersion in the mold. Figure 13 shows that bubble concentration in the mold naturally increases with increasing
volume fraction at the inlet. However, the effect is not
directly proportional.

2. Effect of gas bubble size


Argon bubble size has an important effect that is independent of injection rate. Larger bubbles are predicted
VOLUME 25B, AUGUST 1 9 9 4 - - 5 3 9

0.5 m/s

IJ)IFIIIS//

.........

0.5 m/s

0.5 m/s

0.5 m/s

I,,,,/I!!!!!!!!!!!!!!!!!!!!

~%\~

?Ill~
IIIIt~lt

...........
.............

...........
iJJllllller

......

Ill'trill,

"*ttlttt}

........

Ill~l~t..

...........

llllllllll.

......

*l I I I l | l [

Illll~t,~

...........

IJlllllltl~

......

*llllllll~

Ilillqtt*

.............

IlJl111tltt .....

IIl~lllq

............

.....**jll
.,,..~

Illll~,

ilflllltlll

......

'''tlt|ll~

|ft|r

.....

"'~//IllllJLli

"" ................

1111111111 . . . . . .

,lllllllll

Jl|ll|tttt

.......

fill|ilium

~' . . . . . . . . . . . . . . . . . . .

0,*,.m

Illllllh~t

.l|lllJljl/m

Illltslt~

.......

iiifll|lllJ

. ...................

,...0..

Ililllittl

.,lll/llllllm

......

tllllllt~ ......

......................

#t#JiiIj||j|~

1.5-

..........
""

ttlttttt ........

..flllZlllTitl

)lillq~

.tlflllllll

........

.............

. .........

..................

...................

ua
*l~

,-era
-m

2~

......

.i.,.

11111 I

0
O

,,t. ....................

,t Iltil |

l,lill(

..................

,lttt`
1114

,,.,mm

.. ....
,, ,.,...
,...o, ....

.................

,oo,,*....

2.0.............

. o . , , . . , , ~

(a) Without gas.

(b) Bubble size 3 mm


(11% gas at inlet).

(c) Bubble size 3 mm


(22% gas at inlet).

(d) Bubble size 1 mm


(11% gas at inlet).

Fig. 12--Effect of argon gas bubble size and injection rate on flow pattern in Steel caster.

Mold
Centerlin(

Contour Unit : %

Contour Unit : %

Contour Unit : %

0.5

o
1.0
0
U

1.5

(a) Bubble size 3 mm


(11% gas at inlet).

(b) Bubble size 3 mm


(22% gas at inle0.

(c) Bubble size 1 mm


(11% gas at inlet).

Fig. 13--Effect of argon gas bubble size and injection rate on distribution of gas volume fraction in steel caster.

to leave the m o l d faster and, therefore, to have less influence on the liquid flow pattern. A comparison of
Figures 12(b) and (d), or Figures 13(a) and (c), s h o w s
that smaller bubbles penetrate further across the mold.
The concentration contours s h o w that about 1 in every
10 1 - m m bubbles reaches the narrow face, compared
540--VOLUME

25B,

AUGUST

1994

with less than 1 in 100 for 3 - m m bubbles. This accounts


for the greater influence of small bubbles on the f l o w
pattern. S o m e o f the small bubbles may penetrate into
the lower recirculation z o n e , particularly in narrow
molds. Entrapment deep in the caster is unlikely to present a serious quality problem, h o w e v e r , as it should be
METALLURGICAL

AND

MATERIALS

TRANSACTIONS

extremely rare for large bubbles. The distribution of


small bubbles leaving the top surface is relatively uniform, while large bubbles are concentrated near the
nozzle.

Contour Unit : kW/m 2

Mold
Centefline

i
I

C. Heat Transfer
The effect of argon injection on temperature and
superheat dissipation is shown in Figures 14 through 18
and Table IV. Compared with the fiat temperature distribution across the mold width when there is no gas
(Figure 14(a)), argon bubbles buoy hotter steel to the top
surface, changing its temperature from coldest (1527 ~
at narrow face wall to highest (1531 ~ halfway across
the mold, then to medium (1529 ~ close to the nozzle
wall (Figure 14(b)). Although argon raises the surface
temperature only a few degrees, this small temperature
difference may be very important to solidification at the
meniscus.
Figures 15 through 18 and Table IV show the effect
of argon injection on the "superheat flux" transported
across the interface between the liquid steel and the
solidifying shell against the narrow and wide faces. The
argon gas produces a significant increase in superheat
flux to the upper portion of the wide face, and the meniscus region in particular, where superheat flux is almost doubled. The upward shift in flow causes a greater
portion of the superheat to be removed in the mold and

v.

1.0
('D

P~
1.5

2.0

(a) Without gas.

(b) With gas (bubble size


3 ram, 11% gas at inlet).

Fig. 1 5 - - E f f e c t of argon gas bubbles on heat flux distribution in steel


caster.

Mold
Centerline
-,q
1

0.5

Contour Unit : ~

Oo

# _,,~---',~..~.

0)

.a;

.8

,~"
~"

rco

o,
I

\
\

\
Without Gas
Narrow Face

w,, o.s

W,de Face
w , , . Oas

e~

|
Wide Face
i- Without Gas

30F'' '1J'tl' 2 0 0 J'l '''t'lll'''l'''


400
(a) Without gas.

(b) With gas (bubble size


3 mm, 11% gas at inlet).

Fig. 1 4 - - E f f e c t of argon gas bubbles on temperature distribution in


steel caster.
METALLURGICAL AND MATERIALS TRANSACTIONS B

Heal Flux

600

(kW/m 2)

Fig. 1 6 - - E f f e c t of argon gas bubbles on heat flux along centerlines


of wide and narrow faces (bubble size 3 mm, 11 pet gas at inlet).
VOLUME 25B, AUGUST 1994--541

' ' I ~

'h'

E
(n

I ' ' , 1 , , ,

wmm ~ ~ ~
mDmmw
p gw ~

0.5

~ alp

mpm m~

o
C/I

.o~

~ 1.5

\
11"/o

Without Gas
Gas at Inlet

0
4)

22*/0 Gas at Inlet

r0

_~
0

2.s

t , I ~ J , I , , ~l,

100

200

300

D. Mass Transfer

J , I , J , I i t t I n J

400

500

600

700

Heat Flux ( k W / m 2)
F i g . 1 7 - - E f f e c t o f a r g o n g a s i n j e c t i o n r a t e o n heat flux a l o n g n a r r o w
f a c e c e n t e r l i n e ( b u b b l e size 3 m m ) ,

1,5

-_

.
3

100

200
300 400 500
Heat Flux ~W/m 2)

600

700

F i g . 1 8 - - E f f e c t o f a r g o n g a s b u b b l e size o n h e a t flux a l o n g n a r r o w
f a c e c e n t e r l i n e (11 p c t g a s at inlet).

moves the hottest impingement point further above the


mold exit. The peak superheat flux associated with the
narrow face impingement point is very slightly smaller
than that without gas. These side effects of argon use
could be beneficial in decreasing the potential dangers
related to inadequate liquid superheat at the meniscus
and possibly also thinning of the narrow face shell.

1. Effect of gas injection rate


Increasing argon injection rate amplifies its effect on
heat transfer, in accordance with its effect on the flow
pattern. This is shown in Figure 17 and Table IV. Increasing injection causes a greater upward shift of the
5 4 2 - - V O L U M E 25B, AUGUST 1994

impingement point and a bigger drop of the superheat


flux peak value. It should be noticed that more than
80 pct of the total superheat is transferred to the wide
face for the case with 22 pct 3-mm bubbles (Table IV).
This compares to 70 pct without gas.
2. Effect of gas bubble size
The effect of gas bubble size can be seen in Figure 18
and Table IV. The smaller the bubbles, the stronger their
effect. The narrow face impingement point shifts upward, and the peak value of heat flux decreases systematically with decreasing bubble size (Figure 18). The
total superheat transferred to the shell inside the casting
mold increases from 60 pct (without gas or with 5-mm
bubbles), to 65 pct (with 3-mm bubbles), to 74 pct (with
l-ram bubbles). Heat delivered to the meniscus is also
much greater for the smaller bubbles.

The effect of argon gasbubbles on slab composition


during a steel grade transition was investigated by running the 3-D mass-transfer model, described in
Section V, under the conditions listed as case C in
Table I (both without gas and with 11 pct 3-mm bubbles). A solidification constant, k~heu,of 0.00327 m/s ~
was assumed for the parabolic shell growth function,
which generates a liquid pool length (metallurgical
length) of 19.6 m.
The resulting composition profiles down the solid
slabs are shown in Figure 19. Surface composition
changes slightly when argon is introduced into the casting mold. Argon injection transports new grade to the
meniscus faster, thereby slightly reducing the extent of
intermixing along the slab surface.
There is no measurable change in mixing along the
centerline of the slab, however. The effect of argon gas
on slab composition is smaller than that on superheat
removal because bubbles only affect the very top portion
of the strand (upper 0.5 m). The slab composition depends on mass transfer in the entire liquid pool, which
is more than an order of magnitude longer. The effect
of argon decays very rapidly with distance from the meniscus, and almost completely disappears below about
3 m down the strand, where the solidified shell is only
about 45-mm thick. Thus, internal mixing behavior
below this depth is expected to be unchanged, as observed in Figure 19. Intermixing of grades extends to a
significant distance down the slab centerline and is governed by turbulent diffusion, as discussed and verified
in previous work. t4~
X.

IMPLICATIONS FOR STEEL QUALITY

Flow in the liquid cavity of a continuous steel caster


is greatly affected by argon gas injection. This has important consequences for steel surface quality, through
its effect on surface turbulence, entrapment of bubbles
and inclusions, and superheat delivered to the meniscus
region. This section explores the implications of the
model results on these quality issues.

A. Surface Turbulence
The momentum of fast-floating large (>3 ram) bubbles is likely to generate tremendous surface turbulence
METALLURGICAL AND MATERIALS TRANSACTIONS B

Table IV. Predicted Superheat Distribution


Size 5 ram,
11 Pct Gas
Heat Flow
(kW)
Pct

Without Gas
Heat Flow
(kW)

Superheat Lost to:


Conduction through top
surface flux layer
Convection to shell inside
mold (0 to 0.6 m)
Narrow face
Wide face
Convection to shell just
below mold (0.6 to
1.6 m)
Narrow face
Wide face
Convection to shell
farther below mold
(1.6 to 3.0 m)
Narrow face
Wide face
Dissipation very low in
caster (below 3.0 m)
Total
Superheat into mold
Numerical convergence
errors

_I~II~..~'

I .....

I .....

Pct

-6

-3

Pct

Heat Flow
(kW)

Pct

1.9

11.3

1.9

11.3

1.9

11.3

1.9

99.7
255.1

17.1
43.8

92.3
256.2

15.9
44.0

91.6
290.3

15.7
49.9

91.6
340.7

15.7
58.5

82.1
307.6

14.1
52.9

67.5
118.2

11.6
20.3

56.0
147.3

9.6
25.3

49.2
123.2

8.5
21.2

25.6
117.7

4.4
20.2

30.7
150.5

5.3
25.9

4.8
37.3

0.8
6.4

2.2
31.2

0.4
5.4

2.4
28.3

0.4
4.9

0.3
7.3

0.1
1.3

0.7
12.1

0.1
2.1

0.9
594.8
582.0

0.2
102.2

0.5
597.0
582.0

0.1
102.6

0.5
596.8
582.0

0.1
102.5

0.0
594.5
582.0

0.0
102. I

0.1
595.1
582.0

0.0
102.3

12.8

2.2

15.0

2.6

14.8

2.5

12.5

2.1

13.1

2.3

I .....

I .....

I .....

Gas

Cenlerllne
%

Heat Flow
(kW)

11.3

~SWlthout

Pct

Size 3 mm,
22 Pct Gas

1.9

o..

0.4

Heat Flow
(kW)

Size 1 mm,
11 Pct Gas

11.3

~ 0.6_.-- $ u r t a e ~

Size 3 mm,
11 Pct Gas

With

Gas

12

Distance down Slab (m)


Fig. 1 9 - - E f f e c t of argon bubbles on composition along slab casting
direction during a steel grade transition (bubble size 3 ram, 11 pct
gas at inlet).

when the bubbles explode through the top surface,


thereby presenting a great quality hazard. The predicted
argon distributions (Figure 13) suggest that surface quality problems due to this should be greatest near the central portions of the wide face, where most of the large
bubbles escape. Nonuniform oscillation marks and defects, such as trapped mold slag, which are concentrated
on the central surface of the wide face of the strand could
be evidence of surface turbulence, level fluctuations,
and interrupted liquid powder feeding generated by excessive argon.
To minimize this quality problem, excessive numbers
METALLURGICAL AND MATERIALS TRANSACTIONS B

of large argon bubbles should be avoided. Bubble size


is difficult to control because turbulent fluid flow may
affect the shear forces inside the nozzle walls that govern
the initial bubble size, and may break up or combine
bubbles after their formation. The more practical solution is simply to reduce the argon injection rate to the
minimum possible. Slide gate flow control systems appear to be inherently prone to more quality problems related to excessive gas injection than stopper rod systems,
because of their dependence on argon injection to prevent air aspiration between the plates.
In addition to the increased disturbance of the surface
caused by escaping bubbles, argon injection also affects
surface turbulence through its major influence on the average flow pattern. Quality problems due to surface turbulence have been found to correlate with the speed of
the fluid flowing across the top surface. [48] Lower average surface speeds create fewer level fluctuations and
are less likely to entrain liquid flux.
Increasing argon injection has been shown to direct
increasing flow toward the top surface. High argon levels (o-~0 > 11 pet) create strong flow across the surface
toward the narrow face (Figure 12), so they are detrimental for a second reason. However, small amounts of
argon injection (such as Qg = 0.00005 m 3 / s o r o'go =
5 pet for the casting conditions assumed here) tend to
cancel the general recirculating flow of liquid from the
narrow face back toward the nozzle. The corresponding
decrease in average surface speed may be beneficialJ 27j
An optimum argon injection rate may exist for a given
casting operation. This rate depends on the other factors
that control the flow pattern, such as nozzle geometry
and submergence depth. It also must change with casting
speed, in order to maintain the optimum gas volume
VOLUME 25B, AUGUST 1994--543

fraction. Excessive gas turbulence will occur during


slowdown periods if the argon injection rate is not decreased proportionally. The use of argon may be an important additional parameter to control the flow pattern
in the mold, particularly when casting very wide
slabs. [27]

B. Bubble Entrapment
Pinholes and associated defects such as inclusion clusters form when bubbles that are large enough to cause a
problem (>0.1 mm) are trapped near the meniscus.
Smaller bubbles are more easily entrapped because they
penetrate further into the strand (toward the more stagnant regions), take longer to float, and are more easily
caught between dendrites in the solidifying shell. Results
in Figure 13 show that the greatest meniscus concentration of the critical l-ram bubbles occurs about midway
between the nozzle and the narrow face. This suggests
a slight propensity for this type of defect on this region
of the slab surface.

C. Meniscus Heat Flow


Insufficient heat flow to the meniscus region may be
responsible for quality problems related to deep oscillation marks, nonuniform flow of liquid flux, and subsurface hooks or "extended meniscus" formation. The
latter has been reported to exacerbate the entrapment of
solid inclusions and gas bubbles, as they float upward
near to the solidification front. [481 Argon produces important changes in the dissipation of superheat by bringing hotter steel to the meniscus region. Small bubbles in
particular shift superheat removal higher up the mold
and greatly increase heat flow to the meniscus region.
This should be beneficial for steel quality, in direct proportion to the rate of argon injection.

D. Implications
The results of this work indicate that argon flow rate
should be adjusted according to nozzle geometry and
casting conditions to achieve an optimum balance between bringing hot flow to the surface and avoiding excessive surface turbulence. Considering the other
benefits of argon injection regarding the prevention of
nozzle clogging, enhanced flotation of inclusions, and
slightly sharper grade transitions at the slab surface,
argon can be a valuable asset, if used carefully.
This work has demonstrated that mathematical models
are able to quantify the steady flow pattern found in the
mold cavity, and associated phenomena including the removal of superheat and the effects of argon. In addition,
the behavior of large and small gas bubbles has been
isolated. Surface quality depends greatly on factors such
as surface turbulence, transient flow, inclusion entrapment, mold oscillation, and liquid mold flux behavior,
which are not simulated directly with the present model.
Thus, physical water models and controlled trials on operating steel casters remain important supplements to
understanding flow phenomena. It is important to note
that water model studies can and should account for the
important effects of gas thermal expansion by increasing
the gas injection rate about fivefold (according to
5 4 4 - - V O L U M E 25B, AUGUST 1994

Eq. [20]) to match the bubble fraction entering the mold


cavity of the steel caster.
XI.

CONCLUSIONS

1. Mathematical models have been developed to predict multiphase flow in the continuous slab casting
process and its associated heat and mass transfer.
The models are being applied to understand complex flow-related phenomena in the process.
2. Good agreement has been obtained between the
flow model predictions and observations of flow
pattern characteristics and speed measurements.
Thus, the standard K-e turbulence model appears
able to reasonably reproduce steady-state flow in a
continuous casting mold, even when there is gas injection. Other parts of the model have been validated through comparison with other experimental
data.
3. Both experimental and predicted results show that
there is little difference between the time-averaged
flow patterns from rectangular and round nozzle
ports with similar jet outlet area, jet angle, and casting conditions.
4. Argon gas bubble injection changes the flow pattern
in the upper portion of the mold, shifting the impingement point and recirculation centers upward.
Injection rates of 3 to 5 pct at mold inlet reduce
surface speeds, while rates exceeding 10 pet can reverse the direction of flow toward the narrow face
in the upper region of the mold.
5. The effects of argon gas in a steel caster are intensified, relative to water models, because of the five
times gas volume expansion at high temperature and
the corresponding increase of gas flow rate into the
mold.
6. Increasing gas injection rate or decreasing bubble
size both intensify the changes in the flow pattern.
7. Larger bubbles float more easily and leave the mold
more quickly, so they have less effect on the flow
pattern, but possibly more effect on surface
turbulence.
8. Smaller bubbles penetrate deeper into the liquid
pool, increasing their likelihood of entrapment into
the solidifying shell, causing pinholes and related
inclusion defects.
9. Argon gas injection causes superheat to be removed
higher in the caster, moves the hot spot upward,
lowers the peak superheat flux, and delivers more
heat to the wide face. Both temperature and superheat flux are raised significantly in the meniscus
region.
10. During a steel grade transition, argon injection only
slightly affects slab surface composition and has no
effect on intermixing in the slab interior.
APPENDIX
Gas expansion in casters
Bubble expansion occurs in a steel caster due to the
high temperature of liquid steel. In this work, two simple
models for the expansion of a single bubble heated by
METALLURGICAL AND MATERIALS TRANSACTIONS B

high-temperature ambient fluid were developed, based


on pure conduction and lumped heating with variable
bubble size. These models were applied to estimate the
time needed for expansion of the argon bubbles after
they have been injected into the nozzle.

A. Conduction Model
Assuming that the ambient fluid maintains a constant
bubble surface temperature T=, and ignoring convection
and radiation inside the bubble, the governing equation
and boundary and initial conditions for 1-D transient
heat conduction are as follows:

pgCeg

Ot

r 2 Or

r2k~

[A2]

t=~:T=T~

[A3]

- - - -

The predicted change of bubble size and center temperature are shown in Figure A1 for an initial bubble
size, r~, of 10 ram, initial temperature, T~, of 27 ~ and
ambient fluid temperature, T=, 1550 ~ It can be seen
from Figures AI and A2 that the bubbles are heated very
fast in the ambient fluid. The "heating times" (the time
needed for the bubble to reach 95 pct of its final size
and final temperature) obtained by the conduction and
lumped heating models are 0.15 and 0.02 seconds respectively. This means that bubble heating and expansion are extremely rapid, relative to the mean residence
time of bubbles in the SEN of about 0.5 second for typical casting conditions (case C in Table I with SEN

[A1]

t = 0: T = Ti

OT
r = 0: Or

C. Results

20

'

Lumped

Model

18
a,o

[A41

16

r = r0: T = T~

ro
--

r~

\pg/

[A6]

( ~Ti++ 2 7 3 ~ ' / 3

(pgi)'/~
=

[A5]

273/

Equations [A1] through [A6] are solved by a finitedifference model described elsewhere, t4~ and kg is a
specified function of temperature. ~49,5~

I
I

12

I
I

I
I
L

10

0.0375

pgCpg

dt

ro

t 1/3

ri

\pg/

'

1400

t
T//+ 273/

[A7]
1/3

[AS]

0.15

I-

~ B u b b l e T,
f /Lumped Model,
h=100

1200
o

1000

800

Q.

hb (T - T~)

0.1125

(sec)

Fig. A1--Size change of a single bubble with time when heated in


ambient fluid.

B. Lumped Heating Model

dT

0.075
Time

1600

The lumped heating model assumes that heat transfer


inside the bubble is sufficient to keep a uniform temperature inside the bubble and that the thermal resistance
comes from convection with the ambient fluid. This
model can be written as

Model

14

.~
=
in

where pg = gas density at time t (kg m-3);


Cpg= gas specific heat = 552 (W kg -~ K-~);
k,= gas conductivity at time t ( W m -] K-I);
T,= initial bubble temperature = 27 (~
T~= ambient fluid temperature = 1550 (~
r 0 = bubble radius at time t (m);
r i = initial bubble size (m);
Pgi = initial gas density = 1.6 (kg m-3).

Conduction

E
lid
I-

Bubble Center T,
Conduction Model

!
I
#
#

600
400

200

#
#

These equations are solved for T and r0 using the


fourth-order Runge-Kutta method. The heat transfer
coefficient, hb, for this model was taken to be 100
W/(mZK) from work by Iguchi and co-workers, who fitted hb to experimental data with 30- to 40-ram
bubbles, mu
METALLURGICALAND MATERIALSTRANSACTIONSB

0
0

0.0375

0.075
Time ( s e c )

0.1125

0.15

Fig. A2--Bubble temperature change with time when heated in


ambient fluid.
VOLUME 25B, AUGUST 1994--545

length 0.5 m and SEN inner diameter 76 mmlS'-l). The


lumped model is probably most accurate because the
conduction model neglects the significant amount of
convection inside the bubble and therefore underestimates heating rate. Even the conservative calculation
demonstrates conclusively that bubble heating and expansion are finished inside the SEN. Thus, bubble size
and volume fraction can be calculated by setting bubble
temperature to the liquid steel temperature, as done in
Section III-C of this article.
NOMENCLATURE

C
Deft

Do
Dg

ag
E
F
h
K

k0
Kshell

Lo
th

Lw
N
rl

P
Pro
Prt

Qg
Re

qsh
Sct

To
Tliq

T~ol
T~

Vc
Vgt

Vux
Vgy

relative concentration in strand


specific heat (liquid steel) (J kg -l K - b
effective diffusivity (liquid) (m 2 s -~)
molecular diffusivity (liquid) (m 2 s-t)
apparent gas bubble diffusivity (m: s-~)
diameter of gas bubbles entering mold (mm)
diameter of gas bubbles at injection point in
nozzle (mm)
wall roughness constant (in K-e wall laws)
mass fraction of a given element
heat transfer coefficient (top surface) (W m -2
K-b
turbulent kinetic energy (m 2 s -2)
turbulent kinetic energy (at inlet to mold)
(m 2 s -2)
laminar thermal conductivity (W m -~ K -~)
solidification constant ( m s -~
nominal nozzle submergence depth (from top
surface to top of nozzle port) (m)
inlet height (mm)
inlet width (mm)
jet submergence depth (from top surface to
top of the jet) (m)
strand thickness (across narrow face) (m)
normal direction of boundaries
static pressure (relative to outlet plane of
domain) (N s-:)
laminar Prandtl Number. (C~,l.*oko~)
turbulent Prandtl Number
gas injection rate (entering nozzle gate)
(m 3 S-1)

Reynolds number ( V c ~ p t z o i)
superheat flux from liquid steel to solidifying
shell (W m 2)
turbulent Schmidt number (iz,p-lD-~ I)
temperature (~
casting temperature (pour temperature) (at
inlet to mold) (~
liquidus temperature (~
solidus temperature (~
ambient temperature (~
casting speed ( m s l)
gas bubble terminal velocity (m s J)
gas velocity component in x direction
( m s l)
gas velocity component in y direction

(m s-')
Vx

gas velocity component in z direction (m s -~)


liquid velocity component in x direction
(m s 1)

546--VOLUME 25B, AUGUST 1994

Vy

v,o
V-

liquid normal velocity through inlet to mold


(m s-')
liquid velocity component in y direction
(m s-')
liquid horizontal velocity through inlet to
mold (m s -1)
liquid velocity component in z direction

(m s-')
V_-o

W
Z
Z
O~
OLo

aL
E
E0

JLLeff
~o

P
%
O'go

liquid downward velocity through inlet to


mold (m s -I)
strand width (across wide face) (m)
strand length simulated (m)
distance down strand or slab (m)
jet angle at inlet (Figure 6) (~
nominal angle of nozzle port edges at inlet
(Figure 6) (~
superheat temperature (To - T.q) (~
dissipation rate (m 2 s -3)
inlet dissipation rate (m 2 s -3)
liquid effective viscosity (kg m -I s -~)
liquid laminar (molecular) viscosity
(kg m -l s -l)
liquid turbulent viscosity (kg m -~ s -l)
liquid density (kg m -3)
gas volume fraction (pct)
gas volume fraction (at inlet to mold) (pct)

ACKNOWLEDGMENTS
The authors wish to thank T. Tucker and J. Hester of
Armco Inc. and Y.H. Wang of Inland Steel Corporation
(East Chicago, IN) for their assistance in obtaining measurements on their company water models and for providing data. The steel companies, Armco Inc., Inland
Steel Corporation, LTV Steel (Cleveland, OH), and
BHP Co. Ltd. (Wallsend, Australia), and the National
Science Foundation (Grant No. MSS-8957195) are
gratefully acknowledged for funding which made this research possible. Finally, thanks are due to Fluid
Dynamics Inc. (Evanston, IL) for use of FIPOST and to
the National Center for Supercomputing Applications at
the University of Illinois for time on the CRAY
supercomputers.
REFERENCES
t. M. Satcudean and K.Y.M. Lai: Numer. Heat Transfer, 1988,
vol. 14, pp. 97-111.
2. M. Salcudean, K.Y.M. Lai, and R.I.L. Guthrie: Can. J. Chem.
Eng., 1984, vol. 63, pp. 51-61.
3. M. Salcudean, C.H. Low, A. Hurda, and R.I.L. Guthrie: Chem.
Eng. Commun., 1983, vol. 21, pp. 89-103.
4. D. Mazumdar and R.I.L. Guthrie: Metall. Trans. B, 1985,
vol. 16B, pp. 83-90.
5. T.D. Roy, A.K. Majumdar, and D.B. Spalding: J. Met., t981
(November), pp. 42-47_
6. J. Szekely, N. E1-Kaddah, and J.H. Grevet: in Int. Conf. on
Injection Metallurgy, 1980, pp. 5:1-5:32.
7. J. Szekely, H.J. Wang, and K.M. Kiser: Metall. Trans. B, 1976,
vol. 7B, pp. 287-95.
8. P.E. Anagbo and J.K. Brimacombe: Metall. Trans. B, 1990,
vol. 21B, pp. 637-48.
9. Y.Y. Sheng and G.A. Irons: in Process Technology Conf. (75th
Steelmaking Conf.), L C . Kuhn, ed., The Iron and Steel Society,
Warrendale, PA, 1992, vol. 10.
METALLURGICAL AND MATERIALS TRANSACTIONS B

10. S.T. Johansen and F. Boysan: Metall. Trans. B, 1988, vol. 19B,
pp. 755-64.
11. S.T. Johansen, D.G.C. Robertson, K. Woje, and T.A. Engh:
Metall. Trans. B, 1988, vol. 19B, pp. 745-54.
12. J. Szekely: Fluid Flow Phenomena in Metals Processing,
Academic Press, New York, NY, 1979.
13. R. Kumar and N.R. Kuloor: Adv. Chem. Eng., 1970 vol. 8,
pp. 256-369.
14. J.F. Davidson and O.G. Schiiler: Trans. Inst. Chem. Eng., 1960,
vol. 38, pp. 144-54.
15. T. Asaeda and J. Imberger: J. Fluid Mech., 1993, vol. 249,
pp. 35-57.
16. G.A. Irons and R.I.L. Guthrie: Metall. Trans. B, 1978, vol. 9B,
pp. 101-10.
17. M. Kawakami, S. Hosono, K. Takahashi, and K. Ito: Tetsu-toHagand, 1992, vol. 78 (2), pp. 267-74.
18. P.E. Anagbo, J.K. Brimacombe, and A.E. Wraith: Chem. Eng.
Sci., 1991, vol. 46 (3), pp. 781-88.
19. M. Sano and K. Mori: Trans. Jpn. Inst. Met., 1976, vol. 17,
pp. 344-52.
20. A.H. Castillejos and J.K. Brimacombe: Metatl. Trans. B, 1987,
vol. 18B, pp. 649-58.
21. A.H. Castillejos and J.K. Brimacombe: Metall. Trans. B, 1987,
vol. 18B, pp. 659-71,
22. I. Leibson, E.G. Holcomb, A.G. Cacoso, and J.J, Jacmic:
AIChE. J., 1956, vol. 2 (September), pp. 296-306.
23. R.L. Datta, D.H. Napier, and D.M. Newitt: Trans. Inst. Chem.
Eng., 1950, vol. 28, pp. 14-26.
24. G.C. Vliet and G. Leppert: J. Heat Transfer, Trans. ASME,
1961, May, pp. 163-75.
25. N. Bessho, R. Yoda, and H. Yamasaki: Iron Steelmaker, 1991,
vol. 18 (4), pp. 39-44.
26. N, Tsukamoto, K. Ichikawa, E. lida, A. Morita, and I. Inoue:
in 74th Steelmaking Conf. Proc., P.A. Augius et al., eds., Iron
and Steel Society, Warrendale, PA, 1991, vol. 74, pp. 803-08.
27. P. Andrzejewski, K.-U. Kohler, and W. Pluschkeli: Steel Res.,
1992, vol. 3 (6), pp. 242-46.
28. S.L. Soo: Fluid Dynamics of Multiphase System, Blaisdell
Publishing Co., Waltham, MA, 1967, ch. 3.
29. D.E. Hershey, B.G. Thomas, and F.M. Najjar: Int. J. Numer.
Methods Fluids, 1993, vol. 17 (1), pp. 23-49.
30. B.G. Thomas, L.J. Mika, and F.M. Najjar: Metall. Trans. B,
1990, vol. 21B, pp. 387-400.
31. F.M. Najjar: Master's Thesis, University of Illinois at UrbanaChampaign, Urbana, IL, 1990.

METALLURGICAL AND MATERIALS TRANSACTIONS B

32. B.E. Launder and D.B. Spalding: Comp. Methods Appl. Mech.
Eng., 1974, vol. 3, pp. 269-89.
33. X. Huang, B.G. Thomas, and F.M. Najjar: MetaU. Trans. B,
1992, vol. 23B, pp. 339-56.
34. B.G. Thomas and F.M. Najjar: Appl. Math. Modeling, 1991,
vol. 15, pp. 226-43.
35. X. Huang: Ph.D. Thesis, Tsinghua University, Beijing, 1988.
36. S.V. Patankar: Numerical Heat Transfer and Fluid Flow,
McGraw-Hill, New York, NY, 1980.
37. C. Offerman: Scand. J. Metall., 1981, vol. 10, pp. 25-28.
38. P. Flint: in Proc. of 73rd Steelmaking Conf., Iron and Steel
Society/AIME, Warrendale, PA, 1990, vol. 73, pp. 481-90.
39. B.G. Thomas, W.R. Storkman, and A. Moitra: in Sixth Int. lron
Steel Congress, Iron and Steel Institute of Japan, Tokyo, Japan,
1990, vol. 2, pp. 348-55.
40. X. Huang and B.G. Thomas: Metall. Trans. B, 1993, vol. 24B,
pp. 379-93.
41. M.S. Engelman: FIDAP Theoretical Manual--Revision 4.0,
Fluid Dynamics International, Inc., Evanston, IL, 1986.
42. R. Sussman and J. Schade: Armco Inc., Middletown, OH,
private communication, 1992.
43. R. Sussman, T. Tucker, and J. Hester: Armco Inc., Middletown,
OH, private communication, 1992.
44. J. Birat, M. Larrecq, J. Lamant, and J. Petegnief: in Mold
Operation for Quality and Productivity, Iron and Steel Society,
Warrendale, PA, 1991, pp. 3-14.
45. Y.H. Wang: in Process Technology Conf. (75th Steelmaking
Conf.), L.G. Kuhn, ed., The Iron and Steel Society,
Warrendale, PA, 1992, vol. 10, pp. 271-78.
46. M. Kumada and I. Mabuchi: Trans. Jpn. Soc. Mech. Eng., 1969,
vol. 35, pp. 1053-61.
47. R. Gas: Inland Steel, Inc., East Chicago, IN, private
communication, 1992.
48. J. Watson: Armco Research & Technology, Middletown, OH,
unpublished research, 1993.
49. P.E. Liley: in Handbook of Single-Phase Convective Heat
Transfer, S. Kakac, R.K. Shah, and W. Aung, eds., John Wiley
and Sons, New York, NY, 1987, p. 22.1.
50. R.L. Powell and G.E. Childs: in American Institute of Physics
Handbook, M.W. Zemansky, ed., McGraw-Hill, New York,
NY, 1972, pp. 4-#42--4-204.
51. M. Iguchi, Z.-I. Morita, H. Tokunaga, and H. Tatemichi: Iron
Steel Inst. Jpn. lnt., 1992, vol. 32 (7), pp. 865-72.
52. D.E. Hershey: Master's Thesis, University of Illinois at UrbanaChampaign, Urbana, IL, 1992.

VOLUME 25B, AUGUST 1994--547